Subject: Re: namei caching of newly created files?
To: Alfred Perlstein <>
From: David Laight <>
List: tech-kern
Date: 01/20/2005 21:03:57
> A lot of systems are using "filesystem as a database" semantics
> and just expect something like:
>   user/key -> data
> to scale indefinitely.

I know a company that hit the inode reference count limit, 32766 (or
maybe 65534) subdirectories of a single directory!

I suspect that a linear search may actually be the fastest form
of lookup for 'normal' directories - ie those smaller than a block.


David Laight: